- Continuous Integration System
- Automated Sync from the Tensorflow Repository
- Third Party GitHub Actions
-
See the github workflow files for details on exactly what is run as part of the GitHub Actions CI.
-
Tests can also be run from within a docker container, which can be built with:
docker build -t tflm-ci -f ci/Dockerfile.micro .
or use the tflm-ci docker image from here.
- You will still need to copy or mount your fork of tflite-micro on to this docker container prior to running any tests.
While TfLite Micro and TfLite are in separate GitHub repositories, the two projects continue to share common code.
The TensorFlow repo is the single source of truth for this shared code. As a result, any changes to this shared code must be made in the TensorFlow repo which will then automatically sync'd via a scheduled GitHub workflow.
We use the following third party actions as part of the TFLM continuous integration system.
- Create a PR to automate sync'ing of shared TfLite and TFLM code.